| # | ||||
|---|---|---|---|---|
| 1 | 351 | 575W | $2,2802 used · from $2,195 | |
| 2 | 220 | 450W | $1,2851 used · from $1,290 | |
| 3 | 177.6 | 230W | — | |
| 4 | 126.5 | 70W | $3,3201 used · from $3,280 | |
| 5 | 102.9 | 240W | $8,0501 used · from $7,990 | |
| 6 | 100.6 | 350W | — | |
| 7 | 89.9 | 40W | — | |
| 8 | 85.5 | 45W | $1,9101 used · from $1,890 | |
| 9 | 61.9 | 27W | — | |
| 10 | 50 | 170W | $1621 used · from $165 | |
| 11 | 47 | 145W | — | |
| 12 | 40 | 125W | $781 used · from $75 | |
| 13 | 37 | 60W | — | |
| 14 | 23 | 180W | $3851 used · from $390 | |
| 15 | — | 500W | — | |
| 16 | — | 60W | — | |
| 17 | — | 27W | — | |
| 18 | — | 140W | — | |
| 19 | — | 250W | — | |
| 20 | — | 700W | — |
Hardware we are tracking. Not on the pool yet and no measured pairs.
30 Sep 2026: DeepSeek open-sourced DeepGEMM-Ascend, its matrix kernels for this chip, with the same API as the NVIDIA version. Not on the pool yet: no OpenAI-compatible server uses the kernels, no measured DeepSeek V4 speed exists, and the chip is only offered through Huawei Cloud in China.
